A. Technical Field
The present invention pertains generally to image processing, and relates more particularly to image processing that reduces compression artifacts in multi-layer images.
B. Background of the Invention
Applications that render, record, or display images with limited resources often employ multi-layer compression. In such applications, an input image is first segmented into two or more layers, each of which refers to one type of distinctive characteristic of the input image. These layers and their associated masks are then compressed independently. An advantage of independent compression is that layers containing important information can be preserved while the other layers can undergo more aggressive compression. The set of layers and their associated masks then can form a highly compressed format that retains high information quality.
Segmentation of an input image often results in imprecise definition of the boundary between the foreground layer and the non-foreground layers. This may result in incomplete labeling of the foreground pixels and thus some foreground pixels at the boundary may be labeled non-foreground. These unlabeled foreground pixels may have very different colors than their neighboring non-foreground pixels, and the color differences may lead to strong artifacts in the non-foreground layers after compression. These artifacts will then appear in the images generated from the compressed layers.